GriddedDataset: torch warn fix

[jeffjennings]

Resolves a warning raised when passing a tensor to GriddedDataset.add_mask

[jeffjennings]

I don't know why MyPy is failing in the test, unless it's searching through

        if torch.is_tensor(mask):
            new_2D_mask = mask.clone().detach().to(device)

with a numpy array.

[kadri-nizam]

@jeffjennings

This is a known issue when typing with PyTorch. The recommended way for checking is to check the type explicitly: if isinstance(mask, torch.Tensor). This is the recommended approach by the devs at PyTorch as can be seen from the is_tensor documentation

That being said, my suggestion is to simply skip the check since we have mask.clone().detach() in the proceeding line. Doing new_2D_mask = torch.Tensor(mask).detach().to(device) will perform the same thing for both Tensors and ArrayLike objects.

Made the recommended changes in 30ad09b

[iancze]

Thanks!

[jeffjennings]

Ah ok I see. Thanks for explaining that, @kadri-nizam!